Compliance is Difficult to Scale
By Alex Johnson
#1: Morgan Stanley Has it Backward
Morgan Stanley is testing out GPT-4 to help its human financial advisors:
The experiment is aimed at “helping investment professionals parse through thousands of pages of our in-depth intellectual capital, analyst commentary, and market research in seconds – a process that typically could take more than half an hour,” Morgan Stanley Wealth Management’s head of analytics and data, Jeff McMillan told Fortune. “This will help advisors spend more time focusing on serving their clients.”
The tool will use the technology’s latest version, GPT-4, which launched on Tuesday. It’s currently being tested with 300 advisors, CNBC reported, and when released more broadly, will aid all of Morgan Stanley’s 16,000 advisors.
Morgan Stanley understands that large language models (LLMs) can get stuff wrong:
The tool can sometimes make errors, or “hallucinate,” analysts at investment bank Morgan Stanley said in a note last month. At the time, the analysts wrote that ChatGPT can “generate answers that are seemingly convincing, but are actually wrong.”
So, it’s restricting the use of GPT-4 to summarizing research for human advisors and not letting the chatbot actually interact with clients:
The bank’s use of GPT-4 will be intended to assist human advisors, much like a research assistant would. But it’s not supposed to replace human advisors, who are still needed to interact with clients.
“These things (A.I. tools) don’t have any empathy; they’re just very clever math that is able to regurgitate knowledge,” McMillan told CNBC.
Am I the only one who thinks they’re getting this backward?
I mean, we know that LLMs make factual errors. Even GPT-4. This seems like a bad trait for a research assistant! What are you supposed to do when you have a reasonable belief that something in the summary of all that research is wrong? Go back and double-check everything? Doesn’t that defeat the point?
By contrast, I think we’re underestimating the value that chatbots powered by LLMs can have in facilitating the emotional/social side of customer service tasks like financial advising, as I wrote about a while back:
If these large language models – trained on datasets derived from the entire internet – truly do have some emergent property whereby they are able to communicate using emotions rather than facts, then they might be ideally suited to help consumers reflect on, process, and improve their feelings about money.
#2: The Used Car Market is Bigger Than You Think
A fintech company focused on the automotive industry raised a huge seed round:
Caramel, a company that is transforming private party and independent dealer car sales, today announces the launch of the Caramel checkout platform. The company also announces that it has raised $12 million in seed funding, led by Zeev Ventures, Primera Capital, and Hearst Ventures.
Here’s how Caramel describes itself:
Caramel offers a simple, digital checkout tool to customers and handles the complexities of the sale for them. With Caramel, buyers and sellers can easily and safely transact on any car using any device. Caramel is available in the App Store, Google Play Store, and at drivecaramel.com.
Using Caramel, either party can begin the transaction. Caramel helps customers agree on a price and then authenticates the car and verifies the buyer and seller with a photo ID and selfie. The buyer is offered financing and insurance from top national lenders and carriers, as well as transportation. Caramel securely transfers funds, automates digital paperwork, and handles DMV work for both customers behind the scenes.
The used card market dwarfs the new car market, in terms of the number of transactions. In 2022, nearly 14 million new vehicles were sold in the U.S. For used cars, that number was 36 million, with roughly 20 million of that being sold in private transactions and through independent dealerships.
The addressable market here is huge, and I like the tailwinds for Caramel (continuing shortages of new vehicles, spiraling used car prices, etc.) The big question mark is customer acquisition. Independent dealerships are a pain to sell to, but at least once you have them, you have them. I have no idea how you acquire and retain the long tail of private buyers and sellers in a cost-effective manner. Maybe that’s why they needed $12 million.
#3: Compliance is Difficult to Scale
A banking-as-a-service (BaaS) platform that aspires to be the ‘Unit for emerging markets’ raised a seed round:
Embedded finance solutions in less developed markers are becoming more prominent as platforms look to provide various financial solutions to the unbanked and underserved. Banking infrastructure providers are mainly responsible for the proliferation of such solutions. They allow businesses such as mobile operators, e-commerce platforms and logistics companies to embed and enable banking products for their customers.
Credable, an upstart in this category that provides its clients with the technology stack, scoring capabilities and banking partners, has raised a $2.5 million seed round. It follows the pre-seed round of the embedded finance platform secured in early 2021 and led by The Continent Venture Partners (TCVP).
Credable has gotten impressive traction:
Last May, Credable launched officially with two products: a 30-day term loan product in partnership with Vodacom M-Pesa in Tanzania and a short-term lending product for Diamond Trust Bank in Kenya. Since then, the fintech has enabled over six products for various businesses, from banks and mobile network operators to e-commerce platforms and fintech players across three markets: Tanzania, Kenya and Uganda. So far, over 1.2 million people have opened accounts on its platform and more than 200,000 customers (including consumers and SMEs) have used its banking products. These include savings products, term loans, overdrafts, asset financing and other credit solutions. Credable’s platform has helped disburse $5 million worth of loans disbursed and seen over $3 million of deposits into its savings products, per a statement shared by the startup.
And the theory makes a lot of sense to me – in emerging markets with large unbanked populations, embedded finance can grow really fast because it’s not competing with a lot of legacy financial services companies.
My concern is compliance. A lot of things in BaaS scale quite well (technology, product templates, etc.). Compliance … not so much. Every new country you expand into (Pakistan and Nigeria are reportedly next for Credable) comes with new regulations and new regulators who will need handholding. Individually, each new country might not be a huge lift (emerging markets are generally easier than established ones in this respect), but the maintenance adds up (regulations change and evolve constantly).
I guess the flipside to this concern – and we see this in areas like cross-border payments – is that once you have this multi-country compliance infrastructure built, it’s a hell of a moat.
2 FINTECH CONTENT RECOMMENDATIONS
#1: The AI Revolution Could Be Bigger and Weirder Than We Can Imagine (by Derek Thompson, Plain English) 🎧
Derek Thompson has been doing a fantastic job chronicling the emergence of generative AI and talking through the implications of the technology with (from my perspective) the best mix of realism, wonder, and terror.
This episode focuses on OpenAI’s GPT-4 model. It’s worth a listen.
#2: Secrets of the Magus (by Mark Singer, The New Yorker) 📚
This one has absolutely nothing to do with fintech.
(OK, there is an S&L crisis-era bank failure in the middle of the story … but that’s not the main point, I promise.)
It’s a profile of one of the greatest sleight-of-hand artists in recent history – Ricky Jay. The story is chock-full of great anecdotes about Jay’s exploits (which will, I guarantee you, motivate you to hop on YouTube and see some of his tricks for yourself). It also discusses Jay’s obsessive study of the history and secrets of his field, which is a level of scholarship that we should all aspire to.
It’s a long read, but I enjoyed it.
1 QUESTION TO PONDER
What is the best non-banking/fintech conference or event for someone working in fintech to attend?
I’ve still somehow never been to SXSW, but that one is on my list. What else? Where should we all be going in order to expand our thinking on how we can best solve financial services challenges for customers?